Dulus
An AI agent for your desktop and terminal, on any model you choose. Hosted models run on Fuel, paid in $DULUS.
翻訳待ち
リポジトリ情報はまだ収集していません。
No updates detected yet. GitHub releases will appear here.
Your entire AI universe, mapped. A privacy-first AI usage intelligence platform with interactive WebGL visualizations, analytics, validated imports, and extensible provider integrations.
翻訳待ち
Loading the full README.
Read the full README ↗See how you really use AI, across every tool, without giving away your conversations.
AI Atlas brings ChatGPT, Claude, Gemini, DeepSeek, Kimi and seven more into one privacy-first dashboard, introduced by a scroll-driven WebGL particle universe.
Live demo · Demo dashboard · Quick start · Architecture · Security
Real capture of the landing page. Scroll drives the whole sequence: sphere → orbit → vortex → galaxy.
You probably use three or four AI tools a week. Each one has its own history, its own dashboard and no idea the others exist. Atlas gives you one honest picture.
▶ The full 36-second walkthrough plays below: landing, globe sequence, provider gallery and dashboard.
https://github.com/user-attachments/assets/efd4acf5-ba0b-4ef3-a5e7-496c11170964
Prerequisites: Node.js 22.13+ and npm.
git clone https://github.com/satiricalguru/Ai-atlas.git
cd Ai-atlas
npm ci
npm run dev
Open http://127.0.0.1:5173, then try:
| URL | What you get |
|---|---|
/ | The landing page and particle universe |
/dashboard?mode=demo | The full dashboard with clearly labelled sample data |
/dashboard?mode=local | Your own imports, stored only in this browser |
[!TIP] Neither the demo nor the local workspace needs an account, a database or an API key. Accounts and syncing switch on once you add the Supabase settings described below.
npm run dev and npm run build use the bundled Vinext runtime, which builds a Cloudflare Worker. npm run start previews that build locally. For the standard Next.js runtime (what Vercel uses):
npm run dev:next
npm run build:next
npx next start
| Source | How it gets in |
|---|---|
| ChatGPT | Upload conversations.json from your data export, or connect an OpenAI organization admin key |
| Claude | Upload your conversation export JSON, or connect an Anthropic organization admin key |
| Every other tool | Import normalized CSV or JSON (example below) |
provider,timestamp,model,category,tokens,cost,external_id
chatgpt,2026-10-29T10:00:00Z,gpt-4.1,Writing,1250,0.012,example-1
claude,2026-10-29T11:30:00Z,claude-sonnet-4-5,Coding,2400,0.024,example-2
deepseek,2026-10-30T09:00:00Z,deepseek-chat,Research,,,example-3
Supported provider values: chatgpt · claude · gemini · cursor · copilot · perplexity · deepseek · kimi · grok · mistral · meta · qwen · other.
Imports are validated against a schema, deduplicated by source ID, and capped at 5 MB / 2,000 rows per upload.
flowchart LR
subgraph Sources
A[ChatGPT / Claude exports]
B[CSV / JSON imports]
C[OpenAI & Anthropic<br/>admin usage APIs]
end
A --> V{Zod validation<br/>+ metadata only}
B --> V
C --> I[Inngest durable sync<br/>+ daily sweep]
I --> S[(API snapshots)]
V --> L[(Activity ledger)]
subgraph Supabase
L
S
K[(AES-GCM encrypted<br/>credentials)]
end
L --> D[Dashboard<br/>Overview · Insights · Explorer]
S --> D
D --> F[AI Fingerprint<br/>+ JSON export]
ShaderMaterial animates about 26k points (11k on phones) with simplex-noise breathing, cursor repulsion, a scroll-driven vortex and a sphere-to-galaxy morph. Logo chips are projected from 3D each frame. Lenis provides the smooth scrolling.This is the beginning of the README. Read the rest at the original. Read the full README ↗